All pull requests

Pull requests

• cp: use GitHub App authentication for code freeze

  area:cicomplexity:low
  #4718·svcnemo-autobot opened Oct 18, 2026Contributor·

  ·
  2

• fix(cli): exit cleanly when gym dataset commands hit a config mistake

  community-request
  #4717·sven-rankov opened Oct 18, 2026Contributor·7 tasks done·

  ·
  1

  1

• fix(aviary): sum token usage across every model call in an episode

  area:agentbugcomplexity:lowneeds-reviewQA_fix
  #4716·qiaochuz-nv opened Oct 18, 2026Contributor·

  ·
  1

  2

• docs(agent): organize agent-server pages by reader

  area:agentdocs
  #4713·yaoyu-33 opened Oct 17, 2026Contributor·3/3·5 of 7 tasks·

  ·
  4

• docs(agent): add harness onboarding and upgrade guide

  area:agentdocs
  #4712·yaoyu-33 opened Oct 17, 2026Contributor·2/3·5 of 7 tasks·

  ·
  11

• docs(architecture): describe Environment Server episode orchestration

  area:docsdocs
  #4711·yaoyu-33 opened Oct 17, 2026Contributor·1/3·5 of 7 tasks·

  ·
  6

• fix(sandbox): retry OpenSandbox file transfers that time out

  #4702·ffrujeri opened Oct 16, 2026Contributor·

  ·
  1

• Harden JobBench and WorkspaceBench evals

  area:agentbugcomplexity:highready-to-merge
  #4698·d-molinari opened Oct 16, 2026Contributor·7 tasks done·

  ·
  5

• fix 404 surgeai/GDP.pdf

  area:databugcomplexity:lowneeds-review
  #4697·hsiehjackson opened Oct 16, 2026Contributor·5 tasks done·

  ·
  2

• fix(scicode): export accuracy intervals and per-step stop metadata

  area:evaluationbugcomplexity:lowneeds-review
  #4695·Kipok opened Oct 16, 2026Contributor·

  ·
  1

• fix(google_search): send API key in a header and keep request details out of tool output

  area:environmentbugcomplexity:lowneeds-review
  #4693·ananthsub opened Oct 16, 2026Contributor·

  ·
  2

• fix(resources): move swebench_pro, interactive_browser, and nemo_user_sim onto the session close contract

  #4692·yaoyu-33 opened Oct 16, 2026Contributor·

  ·
  1

• fix(gdpval): follow-ups from the GDPVal/Stirrup decoupling review

  #4691·yaoyu-33 opened Oct 16, 2026Contributor·

  ·
  1

• feat(checkpoint): name the episodes an agent's own /run drives in the agent commit reply

  area:corecomplexity:lowfeatureneeds-review
  #4690·macandro96 opened Oct 16, 2026Contributor·

  ·
  6

• feat(harbor): Compose service builds (M24)

  #4689·adil-a opened Oct 16, 2026Contributor·18/22·

  ·
  2

• feat(conformance): run Stirrup through the harness conformance checks

  #4687·marta-sd opened Oct 16, 2026Contributor·4 of 7 tasks·

  ·
  1

  

• fix(tau2): opt in to policy reasoning replay

  #4686·mpatel31415 opened Oct 16, 2026Contributor·

  ·
  1

• fix(deepswe): score confirmed agent sandbox OOMs as zero

  area:sandboxbugcomplexity:highneeds-review
  #4685·mpatel31415 opened Oct 16, 2026Contributor·

  ·
  16

• docs(training): fix offline SFT/DPO rollout conversion snippets

  area:trainingcomplexity:mediumdocsneeds-review
  #4682·yaoyu-33 opened Oct 16, 2026Contributor·

  ·
  1

  1

• docs(training): document task_source routing instead of agent_ref

  area:trainingcomplexity:mediumdocsneeds-review
  #4681·yaoyu-33 opened Oct 16, 2026Contributor·

  ·
  1

  1

• fix(agent): refuse unsupported agent session seeds with a final 422

  #4680·yaoyu-33 opened Oct 16, 2026Contributor·

  ·
  1

• fix(resources): bound closed sessions and release state on a cookie-less close

  area:env-infrabugcomplexity:highneeds-review
  #4679·yaoyu-33 opened Oct 16, 2026Contributor·

  ·
  1

• docs(evaluate): fix gym eval run examples that fail as written

  area:evaluationcomplexity:lowdocsneeds-review
  #4678·yaoyu-33 opened Oct 16, 2026Contributor·

  ·
  1

  1

• feat(harbor): Compose adapter field coverage (M28)

  #4666·adil-a opened Oct 16, 2026Contributor·20/22·

  ·
  4

• feat(harbor): Compose volumes, bind mounts as uploads, shared and external volumes (M22)

  #4665·adil-a opened Oct 16, 2026Contributor·21/22·

  ·
  4